In the process of RV reducer cycloidal gear modification,it is difficult to comprehensively improve the return difference and transmission error of RV reducer,and most of the cycloidal gear modification methods are single objective modification or weighted single objective modification,which is not in line with the actual engineering situation.Therefore,the transmission error model and return difference model were established by TCA method.A multi-objective particle swarm optimization algorithm with exponential nonlinear decreasing inertia weights and iterative updating learning factors was proposed.The return error and transmission error were taken as objective functions,the modified algorithm was used to optimize the compound modification quantity,and the optimal solution was obtained by membership function.Adams and MATLAB were used to analyze the return difference and transmission error of RV reducer under different modification methods. The results show that the return difference of the final optimized modification is reduced by 0.02′ and 0.04′ respectively,and the transmission error is reduced by 6.6% and 7.8% respectively.